[Mesa-dev] [PATCH 06/19] radv: store the list of attachments for every subpass

This reworks how the depth stencil attachment is used for simplicity. This also introduces radv_render_pass_compile() helper that will be used for further optimizations. Signed-off-by: Samuel Pitoiset --- src/amd/vulkan/radv_cmd_buffer.c | 16 +++--- src/amd/vulkan/radv_meta_clear.c

[Mesa-dev] [PATCH 10/19] radv: add radv_cmd_buffer_end_subpass() helper

To share common code between CmdEndRenderPass() and CmdNextSubpass(). Signed-off-by: Samuel Pitoiset --- src/amd/vulkan/radv_cmd_buffer.c | 70 +--- 1 file changed, 38 insertions(+), 32 deletions(-) diff --git a/src/amd/vulkan/radv_cmd_buffer.c b/src/amd/vulkan

[Mesa-dev] [PATCH 08/19] radv: determine the last subpass id for every attachments

Signed-off-by: Samuel Pitoiset --- src/amd/vulkan/radv_pass.c| 12 src/amd/vulkan/radv_private.h | 3 +++ 2 files changed, 15 insertions(+) diff --git a/src/amd/vulkan/radv_pass.c b/src/amd/vulkan/radv_pass.c index 1102ef689b2..ac9f9381216 100644 --- a/src/amd/vulkan

[Mesa-dev] [PATCH 11/19] radv: move some render pass things to radv_render_pass_compile()

radv_render_pass_compile() is common to vkCreateRenderPass() and vkCreateRenderPass2(). Signed-off-by: Samuel Pitoiset --- src/amd/vulkan/radv_pass.c | 66 ++ 1 file changed, 38 insertions(+), 28 deletions(-) diff --git a/src/amd/vulkan/radv_pass.c b/src/amd

[Mesa-dev] [PATCH 16/19] radv: execute external subpass barriers after ending subpasses

Outgoing dependencies (ie. external) should happen after the subpass. This doesn't change anything for subpass resolves as we already make sure that attachments are shader readable. Signed-off-by: Samuel Pitoiset --- src/amd/vulkan/radv_cmd_buffer.c | 4 ++-- 1 file changed, 2 insertions(

[Mesa-dev] [PATCH 13/19] radv: track if subpasses have color attachments

Signed-off-by: Samuel Pitoiset --- src/amd/vulkan/radv_pass.c | 3 +++ src/amd/vulkan/radv_pipeline.c | 10 +- src/amd/vulkan/radv_private.h | 3 +++ 3 files changed, 7 insertions(+), 9 deletions(-) diff --git a/src/amd/vulkan/radv_pass.c b/src/amd/vulkan/radv_pass.c index

[Mesa-dev] [PATCH 12/19] radv: add radv_render_pass_add_subpass_dep() helper

To share common code that handles subpass dependencies. Signed-off-by: Samuel Pitoiset --- src/amd/vulkan/radv_pass.c | 78 +++--- 1 file changed, 38 insertions(+), 40 deletions(-) diff --git a/src/amd/vulkan/radv_pass.c b/src/amd/vulkan/radv_pass.c index

[Mesa-dev] [PATCH 07/19] radv: use the new attachments array when starting subpasses

Signed-off-by: Samuel Pitoiset --- src/amd/vulkan/radv_cmd_buffer.c | 17 + 1 file changed, 5 insertions(+), 12 deletions(-) diff --git a/src/amd/vulkan/radv_cmd_buffer.c b/src/amd/vulkan/radv_cmd_buffer.c index 440f09a363c..914ff0055d9 100644 --- a/src/amd/vulkan

[Mesa-dev] [PATCH 18/19] radv: do not set preserveAttachments for internal render passes

We don't use that. Signed-off-by: Samuel Pitoiset --- src/amd/vulkan/radv_meta_blit.c | 12 ++-- src/amd/vulkan/radv_meta_blit2d.c | 12 ++-- src/amd/vulkan/radv_meta_clear.c | 8 src/amd/vulkan/radv_meta_resolve_fs.c | 4 ++-- 4 files change

[Mesa-dev] [PATCH 19/19] radv: don't flush src stages when dstStageMask == BOTTOM_OF_PIPE

Original patch by Fredrik Höglund. Signed-off-by: Samuel Pitoiset --- src/amd/vulkan/radv_cmd_buffer.c | 16 +++- src/amd/vulkan/radv_pass.c | 6 -- 2 files changed, 19 insertions(+), 3 deletions(-) diff --git a/src/amd/vulkan/radv_cmd_buffer.c b/src/amd/vulkan

[Mesa-dev] [PATCH 09/19] radv: use the new attachments array in CmdEndRenderPass()

That shouldn't change anything as we check if the last subpass id is the final subpass. Signed-off-by: Samuel Pitoiset --- src/amd/vulkan/radv_cmd_buffer.c | 16 +--- 1 file changed, 13 insertions(+), 3 deletions(-) diff --git a/src/amd/vulkan/radv_cmd_buffer.c b/src/amd/v

[Mesa-dev] [PATCH 17/19] radv: drop useless checks when resolving subpass color attachments

The Vulkan spec says: "If pResolveAttachments is not NULL, for each resolve attachment that does not have the value VK_ATTACHMENT_UNUSED, the corresponding color attachment must not have the value VK_ATTACHMENT_UNUSED." Signed-off-by: Samuel Pitoiset --- src/

[Mesa-dev] [PATCH 14/19] radv: handle subpass dependencies correctly

The different masks should be accumulated. For example if two subpasses declare an outgoing dependency (ie. dst == VK_SUBPASS_EXTERNAL). Signed-off-by: Samuel Pitoiset --- src/amd/vulkan/radv_pass.c | 12 ++-- 1 file changed, 6 insertions(+), 6 deletions(-) diff --git a/src/amd/vulkan

[Mesa-dev] [PATCH 05/19] radv: move subpass image transitions to radv_cmd_buffer_begin_subpass()

Instead of doing them in radv_cmd_buffer_set_subpass(). Signed-off-by: Samuel Pitoiset --- src/amd/vulkan/radv_cmd_buffer.c | 44 +-- src/amd/vulkan/radv_meta_clear.c | 5 +-- src/amd/vulkan/radv_meta_resolve.c| 2 +- src/amd/vulkan/radv_meta_resolve_fs.c

Re: [Mesa-dev] [PATCH 09/19] radv: use the new attachments array in CmdEndRenderPass()

On 1/31/19 11:16 AM, Bas Nieuwenhuizen wrote: On Tue, Jan 29, 2019 at 10:16 PM Samuel Pitoiset wrote: That shouldn't change anything as we check if the last subpass id is the final subpass. Signed-off-by: Samuel Pitoiset --- src/amd/vulkan/radv_cmd_buffer.c | 16 +--- 1

[Mesa-dev] [PATCH v2 1/4] radv: gather more info about push constants

This is needed in order to inline some push constants when possible. This also adds a new helper for initializing the pass. v2: - use MIN2/MAX2 - use UINT8_MAX Signed-off-by: Samuel Pitoiset --- src/amd/vulkan/radv_nir_to_llvm.c | 2 ++ src/amd/vulkan/radv_private.h | 2 ++ src/amd

Re: [Mesa-dev] [PATCH] ac/radv/radeonsi: add ac_get_num_physical_sgprs() helper

Reviewed-by: Samuel Pitoiset On 2/1/19 11:28 AM, Timothy Arceri wrote: --- src/amd/common/ac_gpu_info.h | 6 ++ src/amd/vulkan/radv_device.c | 2 +- src/amd/vulkan/radv_shader.c | 4 ++-- src/amd/vulkan/radv_shader.h | 6 -- src

Re: [Mesa-dev] [PATCH] radv: take LDS into account for compute shader occupancy stats

Fixed. On 2/1/19 2:57 PM, Mike Lothian wrote: Hi I think you've left a few references to the old ac_nir_get_max_workgroup_size FAILED: src/amd/vulkan/9198681@@vulkan_radeon@sha/radv_nir_to_llvm.c.o x86_64-pc-linux-gnu-gcc -m32 -Isrc/amd/vulkan/9198681@@vulkan_radeon@sha -Isrc/amd/vulkan -I../m

Re: [Mesa-dev] [RFC PATCH 4/4] radv: add support for push constants inlining when possible

On 2/5/19 10:58 AM, Bas Nieuwenhuizen wrote: On Fri, Jan 25, 2019 at 5:27 PM Samuel Pitoiset wrote: This removes some scalar loads from shaders, but it increases the number of SET_SH_REG packets. This is currently basic but it could be improved if needed. Inlining dynamic offsets might also

Re: [Mesa-dev] [RFC PATCH 4/4] radv: add support for push constants inlining when possible

On 2/5/19 11:29 AM, Bas Nieuwenhuizen wrote: On Tue, Feb 5, 2019 at 11:07 AM Samuel Pitoiset wrote: On 2/5/19 10:58 AM, Bas Nieuwenhuizen wrote: On Fri, Jan 25, 2019 at 5:27 PM Samuel Pitoiset wrote: This removes some scalar loads from shaders, but it increases the number of SET_SH_REG

[Mesa-dev] [PATCH] nir: compute new range when shrinking push constants loads

RADV will rely on that range for inlining push constants. Signed-off-by: Samuel Pitoiset --- src/compiler/nir/nir_opt_shrink_load.c | 5 + 1 file changed, 5 insertions(+) diff --git a/src/compiler/nir/nir_opt_shrink_load.c b/src/compiler/nir/nir_opt_shrink_load.c index f97b7f9b67f

[Mesa-dev] [PATCH v3 2/4] radv: gather if shaders load dynamic offsets separately

Signed-off-by: Samuel Pitoiset --- src/amd/vulkan/radv_shader.h | 1 + src/amd/vulkan/radv_shader_info.c | 1 + 2 files changed, 2 insertions(+) diff --git a/src/amd/vulkan/radv_shader.h b/src/amd/vulkan/radv_shader.h index 92886188459..c194401c02d 100644 --- a/src/amd/vulkan/radv_shader.h

[Mesa-dev] [PATCH v3 1/4] radv: gather more info about push constants

: Samuel Pitoiset fix gather --- src/amd/vulkan/radv_nir_to_llvm.c | 2 ++ src/amd/vulkan/radv_private.h | 2 ++ src/amd/vulkan/radv_shader.h | 4 src/amd/vulkan/radv_shader_info.c | 37 ++- 4 files changed, 44 insertions(+), 1 deletion(-) diff --git a

[Mesa-dev] [PATCH v3 3/4] radv: keep track of the number of remaining user SGPRs

Signed-off-by: Samuel Pitoiset --- src/amd/vulkan/radv_nir_to_llvm.c | 4 1 file changed, 4 insertions(+) diff --git a/src/amd/vulkan/radv_nir_to_llvm.c b/src/amd/vulkan/radv_nir_to_llvm.c index dbeae2aafde..29300aeab9f 100644 --- a/src/amd/vulkan/radv_nir_to_llvm.c +++ b/src/amd/vulkan

[Mesa-dev] [PATCH v3 4/4] radv: add support for push constants inlining when possible

: 1442 -> 1429 (-0.90 %) Code Size: 8126688 -> 7940960 (-2.29 %) bytes Max Waves: 80952 -> 81645 (0.86 %) v2: - check has_only_32bit_push_constants - handle base != 0 Signed-off-by: Samuel Pitoiset --- src/amd/common/ac_nir_to_llvm.c | 27 +-- src/amd/common/ac_shader_ab

[Mesa-dev] [PATCH] radv: fix compiler issues with GCC 9

nd of containing function. This has been fixed in GCC 9. Code that relied on this extended lifetime needs to be fixed, move the compound literals to whatever scope they need to accessible in." Bugzilla: https://bugs.freedesktop.org/show_bug.cgi?id=109543 Cc: Signed-off-by: Samuel Pitoiset --

[Mesa-dev] [PATCH] radv: fix using LOAD_CONTEXT_REG with old GFX ME firmwares on GFX8

This fixes a critical issue. Cc: Bugzilla: https://bugs.freedesktop.org/show_bug.cgi?id=109575 Signed-off-by: Samuel Pitoiset --- src/amd/vulkan/radv_cmd_buffer.c | 5 ++--- src/amd/vulkan/radv_device.c | 5 + src/amd/vulkan/radv_private.h| 3 +++ 3 files changed, 10 insertions

[Mesa-dev] [PATCH v2] radv: fix using LOAD_CONTEXT_REG with old GFX ME firmwares on GFX8

This fixes a critical issue. v2: - use feature, not version Cc: Bugzilla: https://bugs.freedesktop.org/show_bug.cgi?id=109575 Signed-off-by: Samuel Pitoiset --- src/amd/vulkan/radv_cmd_buffer.c | 5 ++--- src/amd/vulkan/radv_device.c | 5 + src/amd/vulkan/radv_private.h| 3 +++ 3

[Mesa-dev] [PATCH 3/4] radv: store vertex attribute formats as pipeline keys

The formats will be used for reducing the number of loaded channels. Signed-off-by: Samuel Pitoiset --- src/amd/vulkan/radv_pipeline.c | 22 +++--- src/amd/vulkan/radv_private.h | 1 + src/amd/vulkan/radv_shader.h | 1 + 3 files changed, 21 insertions(+), 3 deletions

[Mesa-dev] [PATCH 1/4] ac: make use of ac_build_expand_to_vec4() in visit_image_store()

And make ac_build_expand() a static function. Signed-off-by: Samuel Pitoiset --- src/amd/common/ac_llvm_build.c | 9 + src/amd/common/ac_llvm_build.h | 3 --- src/amd/common/ac_nir_to_llvm.c | 2 +- 3 files changed, 6 insertions(+), 8 deletions(-) diff --git a/src/amd/common

[Mesa-dev] [PATCH 4/4] radv: reduce the number of loaded channels for vertex input fetches

4 -> 15984 (3.30 %) This mostly helps Croteam games (Talos/Sam2017). Signed-off-by: Samuel Pitoiset --- src/amd/vulkan/radv_nir_to_llvm.c | 83 ++- 1 file changed, 81 insertions(+), 2 deletions(-) diff --git a/src/amd/vulkan/radv_nir_to_llvm.c b/src/amd/vulkan/r

Re: [Mesa-dev] [PATCH 00/38] radv, ac: 16-bit and 8-bit arithmetic and 8-bit storage

What's the status of this? On 12/7/18 6:21 PM, Rhys Perry wrote: This series add support for: - VK_KHR_shader_float16_int8 - VK_AMD_gpu_shader_half_float - VK_AMD_gpu_shader_int16 - VK_KHR_8bit_storage on VI+. Half floats are currently disabled on LLVM 7 because of a bug causing large memory usa

Re: [Mesa-dev] [PATCH 00/38] radv, ac: 16-bit and 8-bit arithmetic and 8-bit storage

RC enabling SLP vectorizer also uncovered a RA bug with a shader. I think I'll look into the issues with patch 35 again. On Tue, 12 Feb 2019 at 16:30, Samuel Pitoiset wrote: What's the status of this? On 12/7/18 6:21 PM, Rhys Perry wrote: This series add support for: - V

[Mesa-dev] [PATCH] radv: always export gl_SampleMask when the fragment shader uses it

For some reasons, this breaks trees rendering in Project Cars. Fixes: 85010585cde ("radv: only enable gl_SampleMask if MSAA is enabled too") Bugzilla: https://bugs.freedesktop.org/show_bug.cgi?id=109401 Signed-off-by: Samuel Pitoiset --- src/amd/vulkan/radv_pipeline.c | 8 --

[Mesa-dev] [PATCH 2/4] radv: use MAX_{VBS, VERTEX_ATTRIBS} when defining max vertex input limits

Signed-off-by: Samuel Pitoiset --- src/amd/vulkan/radv_device.c | 4 ++-- 1 file changed, 2 insertions(+), 2 deletions(-) diff --git a/src/amd/vulkan/radv_device.c b/src/amd/vulkan/radv_device.c index 0fef92773e1..9778b13ce86 100644 --- a/src/amd/vulkan/radv_device.c +++ b/src/amd/vulkan

[Mesa-dev] [PATCH] radv/winsys: fix BO list creation when RADV_DEBUG=allbos is set

Fixes: 50fd253bd6e ("radv/winsys: Add priority handling during submit.") Signed-off-by: Samuel Pitoiset --- src/amd/vulkan/winsys/amdgpu/radv_amdgpu_cs.c | 1 + 1 file changed, 1 insertion(+) diff --git a/src/amd/vulkan/winsys/amdgpu/radv_amdgpu_cs.c b/src/amd/vulkan/win

Re: [Mesa-dev] [PATCH 00/38] radv, ac: 16-bit and 8-bit arithmetic and 8-bit storage

ase and handle Marek feedbacks, at least? I will review the v2. Thanks Rhys. On Tue, 12 Feb 2019 at 17:08, Samuel Pitoiset wrote: How about splitting this series in four different parts? One for every extension? Is this doable without too much troubles? On 2/12/19 6:02 PM, Rhys Perry wrote: I

Re: [Mesa-dev] [PATCH 4/4] radv: reduce the number of loaded channels for vertex input fetches

On 2/13/19 10:59 PM, Bas Nieuwenhuizen wrote: On Tue, Feb 12, 2019 at 3:07 PM Samuel Pitoiset wrote: It's unnecessary to load more channels than the vertex attribute format. The remaining channels are filled with 0 for y and z, and 1 for w. 29077 shaders in 15096 tests Totals: SGPRS: 13

[Mesa-dev] [PATCH 2/2] ac: use new LLVM 8 intrinsic when loading 16-bit values

Signed-off-by: Samuel Pitoiset --- src/amd/common/ac_llvm_build.c | 41 ++ 1 file changed, 27 insertions(+), 14 deletions(-) diff --git a/src/amd/common/ac_llvm_build.c b/src/amd/common/ac_llvm_build.c index 3acf41728ac..867a13622f9 100644 --- a/src/amd/common

[Mesa-dev] [PATCH 1/2] ac: add ac_build_llvm8_tbuffer_load() helper

It uses the new LLVM intrinsics. Signed-off-by: Samuel Pitoiset --- src/amd/common/ac_llvm_build.c | 38 ++ src/amd/common/ac_llvm_build.h | 14 + 2 files changed, 52 insertions(+) diff --git a/src/amd/common/ac_llvm_build.c b/src/amd/common

[Mesa-dev] [PATCH] radv: fix invalid element type when filling vertex input default values

The elements added into a vector should have the same type as the first one, otherwise this hits an assertion in LLVM. Fixes: 4b3549c0846 ("radv: reduce the number of loaded channels for vertex input fetches") reported-by: Philip Rebohle Signed-off-by: Samuel Pitoiset --- src/

[Mesa-dev] [PATCH] radv: write the alpha channel of MRT0 when alpha coverage is enabled

Bugzilla: https://bugs.freedesktop.org/show_bug.cgi?id=109597 Cc: 18.3 19.0 Signed-off-by: Samuel Pitoiset --- src/amd/vulkan/radv_pipeline.c | 8 1 file changed, 8 insertions(+) diff --git a/src/amd/vulkan/radv_pipeline.c b/src/amd/vulkan/radv_pipeline.c index 9745a1f2aa7

Re: [Mesa-dev] [PATCH v2 02/41] radv: ensure export arguments are always float

Patch 1-2 are: Reviewed-by: Samuel Pitoiset On 2/16/19 1:21 AM, Rhys Perry wrote: So that the signature is correct and consistent, the inputs to a export intrinsic should always be 32-bit floats. This and the previous commit fixes a large amount crashes from dEQP

Re: [Mesa-dev] [PATCH 00/38] radv, ac: 16-bit and 8-bit arithmetic and 8-bit storage

On 2/16/19 1:21 AM, Rhys Perry wrote: This series add support for: - VK_KHR_shader_float16_int8 - VK_AMD_gpu_shader_half_float - VK_AMD_gpu_shader_int16 - VK_KHR_8bit_storage on VI+. Half floats are disabled on LLVM 7 because of a bug causing large memory usage and long (or unbounded) compilatio

Re: [Mesa-dev] [PATCH v2 06/41] ac/nir: fix 16-bit ssbo stores

Does this fix anything know? There is a 16-bit version of tbuffer.store, maybe we should use it? On 2/16/19 1:21 AM, Rhys Perry wrote: Signed-off-by: Rhys Perry --- src/amd/common/ac_nir_to_llvm.c | 2 ++ 1 file changed, 2 insertions(+) diff --git a/src/amd/common/ac_nir_to_llvm.c b/src/am

Re: [Mesa-dev] [PATCH v2 07/41] ac/nir: implement 8-bit nir_load_const_instr

Reviewed-by: Samuel Pitoiset On 2/16/19 1:21 AM, Rhys Perry wrote: Signed-off-by: Rhys Perry --- src/amd/common/ac_nir_to_llvm.c | 4 1 file changed, 4 insertions(+) diff --git a/src/amd/common/ac_nir_to_llvm.c b/src/amd/common/ac_nir_to_llvm.c index b260142c177..f39232b91a1 100644

Re: [Mesa-dev] [PATCH v2 09/41] ac/nir: fix 64-bit nir_op_f2f16_rtz

Reviewed-by: Samuel Pitoiset On 2/16/19 1:21 AM, Rhys Perry wrote: Signed-off-by: Rhys Perry --- src/amd/common/ac_nir_to_llvm.c | 2 ++ 1 file changed, 2 insertions(+) diff --git a/src/amd/common/ac_nir_to_llvm.c b/src/amd/common/ac_nir_to_llvm.c index 691d444db05..741059b5f1a 100644

Re: [Mesa-dev] [PATCH v2 10/41] ac/nir: make ac_build_clamp work on all bit sizes

We usually use 'name' instead of 'intr'. With that renamed, patch is: Reviewed-by: Samuel Pitoiset On 2/16/19 1:21 AM, Rhys Perry wrote: v2: don't use ac_get_zerof() and ac_get_onef() Signed-off-by: Rhys Perry --- src/amd/common/ac_llvm_build.c | 13

Re: [Mesa-dev] [PATCH v2 12/41] ac/nir: make ac_build_isign work on all bit sizes

Reviewed-by: Samuel Pitoiset On 2/16/19 1:22 AM, Rhys Perry wrote: v2: don't use ac_get_zero(), ac_get_one() and ac_int_of_size() Signed-off-by: Rhys Perry --- src/amd/common/ac_llvm_build.c | 27 --- 1 file changed, 4 insertions(+), 23 deletions(-) diff --

Re: [Mesa-dev] [PATCH v2 17/41] ac/nir: implement half-float nir_op_ldexp

Patches 14-17 are: Reviewed-by: Samuel Pitoiset On 2/16/19 1:22 AM, Rhys Perry wrote: Signed-off-by: Rhys Perry --- src/amd/common/ac_nir_to_llvm.c | 4 +++- 1 file changed, 3 insertions(+), 1 deletion(-) diff --git a/src/amd/common/ac_nir_to_llvm.c b/src/amd/common/ac_nir_to_llvm.c

Re: [Mesa-dev] [PATCH v2 18/41] radv: lower 16-bit flrp

Reviewed-by: Samuel Pitoiset On 2/16/19 1:22 AM, Rhys Perry wrote: Signed-off-by: Rhys Perry --- src/amd/vulkan/radv_shader.c | 1 + 1 file changed, 1 insertion(+) diff --git a/src/amd/vulkan/radv_shader.c b/src/amd/vulkan/radv_shader.c index 1dcb0606246..adba730ad8b 100644 --- a/src/amd

[Mesa-dev] [PATCH] radv: fix clearing attachments in secondary command buffers

If no framebuffer is bound, get the number of samples and the image format from the render pass. This fixes new CTS dEQP-VK.geometry.layered.*.secondary_cmd_buffer. Cc: 18.3 19.0 Signed-off-by: Samuel Pitoiset --- src/amd/vulkan/radv_meta_clear.c | 53 ++-- 1 file

[Mesa-dev] [PATCH 1/2] radv: fix out-of-bounds access when copying descriptors BO list

We shouldn't increment the buffer list pointers twice. This fixes some crashes with new CTS dEQP-VK.binding_model.descriptor_copy.*. Cc: 18.3 19.0 Signed-off-by: Samuel Pitoiset --- src/amd/vulkan/radv_descriptor_set.c | 2 -- 1 file changed, 2 deletions(-) diff --git a/src/amd/v

[Mesa-dev] [PATCH 2/2] radv: don't copy buffer descriptors list for samplers

Sampler descriptors don't have a buffer list. This fixes some crashes with new CTS dEQP-VK.binding_model.descriptor_copy.*.sampler_*. Cc: 18.3 19.0 Signed-off-by: Samuel Pitoiset --- src/amd/vulkan/radv_descriptor_set.c | 6 +- 1 file changed, 5 insertions(+), 1 deletion(-) diff --

Re: [Mesa-dev] [PATCH v2 06/41] ac/nir: fix 16-bit ssbo stores

Reviewed-by: Samuel Pitoiset On 2/18/19 2:23 PM, Rhys Perry wrote: I don't see a 16-bit version of tbuffer.store in IntrinsicsAMDGPU.td and simply changing "llvm.amdgcn.tbuffer.store.i32" to "llvm.amdgcn.tbuffer.store.i16" and removing the zext doesn't seem to wo

[Mesa-dev] [PATCH 2/3] radv: store more vertex attribute infos as pipeline keys

They are required for using typed buffer loads. Signed-off-by: Samuel Pitoiset --- src/amd/vulkan/radv_pipeline.c | 27 +++ src/amd/vulkan/radv_private.h | 4 src/amd/vulkan/radv_shader.h | 6 ++ 3 files changed, 37 insertions(+) diff --git a/src/amd

[Mesa-dev] [PATCH 1/3] ac: rework typed buffers loads for LLVM 7

Be more generic, this will be used by an upcoming series. Signed-off-by: Samuel Pitoiset --- src/amd/common/ac_llvm_build.c | 106 src/amd/common/ac_llvm_build.h | 30 - src/amd/common/ac_nir_to_llvm.c | 4 +- 3 files changed, 83 insertions(+), 57

[Mesa-dev] [PATCH 3/3] radv: use typed buffer loads for vertex input fetches

Waves: 86460 -> 86759 (0.35 %) This gives a really tiny boost. Signed-off-by: Samuel Pitoiset --- src/amd/vulkan/radv_cmd_buffer.c | 21 +- src/amd/vulkan/radv_nir_to_llvm.c | 47 +-- src/amd/vulkan/radv_pipeline.c| 37 ++-- src/amd/

[Mesa-dev] [PATCH] radv: only load 2-dwords for vertex buffers when robustness is disabled

-> 120509 (0.43 %) Signed-off-by: Samuel Pitoiset --- src/amd/vulkan/radv_device.c | 2 ++ src/amd/vulkan/radv_nir_to_llvm.c | 21 - src/amd/vulkan/radv_private.h | 1 + src/amd/vulkan/radv_shader.c | 1 + src/amd/vulkan/radv_shader.h | 1 + 5

[Mesa-dev] [PATCH] rav: use 32_AR instead of 32_ABGR when alpha coverage is required

This export format is faster. Seems to improve performance in Wreckfest. Signed-off-by: Samuel Pitoiset --- src/amd/vulkan/radv_pipeline.c | 2 +- 1 file changed, 1 insertion(+), 1 deletion(-) diff --git a/src/amd/vulkan/radv_pipeline.c b/src/amd/vulkan/radv_pipeline.c index 30c3f60790e

[Mesa-dev] [PATCH] radv: properly align the fence and EOP bug VA on GFX9

If alignement is 0, offets returned by radv_cmd_buffer_upload_alloc() are always 0. These two virtual addresses were pointing at the same location. Cc: 18.3 19.0 Signed-off-by: Samuel Pitoiset --- src/amd/vulkan/radv_cmd_buffer.c | 5 +++-- 1 file changed, 3 insertions(+), 2 deletions(-) diff

[Mesa-dev] [PATCH v2] radv: properly align the fence and EOP bug VA on GFX9

If alignement is 0, offets returned by radv_cmd_buffer_upload_alloc() are always 0. These two virtual addresses were pointing at the same location. v2: - add an asertion that checks if alignment is power of two Cc: 18.3 19.0 Signed-off-by: Samuel Pitoiset --- src/amd/vulkan/radv_cmd_buffer.c

[Mesa-dev] [PATCH] ac/nir: implement emit_{imul, umul}_2x32_64 opcodes

Fixes: 58bcebd987b ("spirv: Allow [i/u]mulExtended to use new nir opcode") Signed-off-by: Samuel Pitoiset --- src/amd/common/ac_nir_to_llvm.c | 36 + 1 file changed, 36 insertions(+) diff --git a/src/amd/common/ac_nir_to_llvm.c b/src/amd/common/ac_nir

[Mesa-dev] [PATCH] radv: allocate enough space in cmdbuf when starting a subpass

ed-off-by: Samuel Pitoiset --- src/amd/vulkan/radv_cmd_buffer.c | 2 +- 1 file changed, 1 insertion(+), 1 deletion(-) diff --git a/src/amd/vulkan/radv_cmd_buffer.c b/src/amd/vulkan/radv_cmd_buffer.c index ad0b934ddfc..3e652018499 100644 --- a/src/amd/vulkan/radv_cmd_buffer.c +++ b/src/amd/

Re: [Mesa-dev] [PATCH] radv: allocate enough space in cmdbuf when starting a subpass

On 3/5/19 2:34 PM, Bas Nieuwenhuizen wrote: On Tue, Mar 5, 2019 at 10:42 AM Samuel Pitoiset wrote: This fixes some CTS crashes with: dEQP-VK.renderpass2.suballocation.attachment_write_mask.attachment_count_8.start_index_* Ideally, we should check cmd_buffer->cs->max_dw because th

Re: [Mesa-dev] [PATCH] ac/nir: implement emit_{imul, umul}_2x32_64 opcodes

On 3/5/19 2:01 PM, Bas Nieuwenhuizen wrote: On Tue, Mar 5, 2019 at 10:30 AM Samuel Pitoiset wrote: Fixes: 58bcebd987b ("spirv: Allow [i/u]mulExtended to use new nir opcode") Signed-off-by: Samuel Pitoiset --- src/amd/common/ac_nir_to_llvm.c | 36

[Mesa-dev] [PATCH v2] ac/nir: implement emit_{imul, umul}_2x32_64 opcodes

v2: - remove dead variables Fixes: 58bcebd987b ("spirv: Allow [i/u]mulExtended to use new nir opcode") Signed-off-by: Samuel Pitoiset --- src/amd/common/ac_nir_to_llvm.c | 34 + 1 file changed, 34 insertions(+) diff --git a/src/amd/common/ac_nir_to_ll

[Mesa-dev] [PATCH] radv: fix binding transform feedback buffers

The mask should be accumulated if two calls are used for binding two buffers at different indexes. Otherwise, the driver only accounts for the last one. Noticed while glancing at this code. Cc: 18.3 19.0 Signed-off-by: Samuel Pitoiset --- src/amd/vulkan/radv_cmd_buffer.c | 2 +- 1 file

[Mesa-dev] [PATCH] radv: fix color conversions for normalized uint/sint formats

ally disabled). Cc: 18.3 19.0 Signed-off-by: Samuel Pitoiset --- src/amd/vulkan/radv_formats.c | 20 1 file changed, 16 insertions(+), 4 deletions(-) diff --git a/src/amd/vulkan/radv_formats.c b/src/amd/vulkan/radv_formats.c index 0a3ff9ebbd9..9c61e769ebd 100644 --- a/src

[Mesa-dev] [PATCH] radv: disable DCC for X4 Foundations to workaround a GPU hang

The game apparently hangs inside a copy image operation, but only when DCC is enabled. I haven't figured out the root cause yet, but this workaround fixes the problem and allows people to play that title, at least. Cc: 18.3 19.0 Signed-off-by: Samuel Pitoiset --- src/amd/vulkan/radv_dev

[Mesa-dev] [PATCH] radv: increase aligment when allocating descriptors for meta operations

19.0 Signed-off-by: Samuel Pitoiset --- src/amd/vulkan/radv_cmd_buffer.c | 2 +- 1 file changed, 1 insertion(+), 1 deletion(-) diff --git a/src/amd/vulkan/radv_cmd_buffer.c b/src/amd/vulkan/radv_cmd_buffer.c index 5b66930d137..24ed6d47a51 100644 --- a/src/amd/vulkan/radv_cmd_buffer.c +++ b/src/amd

Re: [Mesa-dev] [PATCH] radv: disable DCC for X4 Foundations to workaround a GPU hang

Superseded by https://patchwork.freedesktop.org/patch/290846/?series=57689&rev=1 On 3/6/19 7:39 PM, Samuel Pitoiset wrote: The game apparently hangs inside a copy image operation, but only when DCC is enabled. I haven't figured out the root cause yet, but this workaround fixes the pr

[Mesa-dev] [PATCH] Revert "radv: execute external subpass barriers after ending subpasses"

This changes is actually wrong because we have to sync before doing image layout transitions. This fixes rendering issues in Batman, Path of Exile and probably more titles. This reverts commit 76c17cfd8da017ebd19be33ba6cef888957a6758. Cc: 19.0 Signed-off-by: Samuel Pitoiset --- src/amd

[Mesa-dev] [PATCH] radv: fix pointSizeRange limits

The values should match the ones that are emitted. This fixes new CTS dEQP-VK.rasterization.primitive_size.points.*. Fixes: f4e499ec791 ("radv: add initial non-conformant radv vulkan driver") Signed-off-by: Samuel Pitoiset --- src/amd/vulkan/radv_device.c | 2 +- 1 file changed, 1

Re: [Mesa-dev] [PATCH] radv: fix binding transform feedback buffers

should just check if buffers aren't VK_NULL_HANDLE? Reviewed-by: Bas Nieuwenhuizen On Tue, Mar 5, 2019 at 6:06 PM Samuel Pitoiset wrote: The mask should be accumulated if two calls are used for binding two buffers at different indexes. Otherwise, the driver only accounts for the las

Re: [Mesa-dev] [PATCH] radv: Fix driverUUID

Reviewed-by: Samuel Pitoiset On 3/12/19 4:07 PM, Józef Kucia wrote: --- src/amd/vulkan/radv_device.c | 2 +- 1 file changed, 1 insertion(+), 1 deletion(-) diff --git a/src/amd/vulkan/radv_device.c b/src/amd/vulkan/radv_device.c index 83d218fb6bf0..6deb69d22d48 100644 --- a/src/amd/vulkan

[Mesa-dev] [PATCH 00/11] ac: use LLVM 8 buffer intrinsics everywhere

Hi, This small series makes use of new LLVM 8 buffer intrinsics. No CTS regressions on GFX8 with LLVM 7, 8 and master. Please review, Thanks! Samuel Pitoiset (11): ac: fix glc parameter use for new LLVM 8 typed buffer intrinsics ac: make use of ac_get_store_intr_attribs() where possible

[Mesa-dev] [PATCH 06/11] ac/nir: use new LLVM 8 intrinsics for SSBO atomic operations

Signed-off-by: Samuel Pitoiset --- src/amd/common/ac_nir_to_llvm.c | 65 + 1 file changed, 41 insertions(+), 24 deletions(-) diff --git a/src/amd/common/ac_nir_to_llvm.c b/src/amd/common/ac_nir_to_llvm.c index 55d3ce90ce4..922f1063c79 100644 --- a/src/amd/common

[Mesa-dev] [PATCH 03/11] ac/nir: set attrib flags for SSBO and image store operations

For consistency regarding other store operations. Signed-off-by: Samuel Pitoiset --- src/amd/common/ac_nir_to_llvm.c | 9 ++--- 1 file changed, 6 insertions(+), 3 deletions(-) diff --git a/src/amd/common/ac_nir_to_llvm.c b/src/amd/common/ac_nir_to_llvm.c index a7b3fdf64aa..ff29345ffe5

[Mesa-dev] [PATCH 02/11] ac: make use of ac_get_store_intr_attribs() where possible

Signed-off-by: Samuel Pitoiset --- src/amd/common/ac_llvm_build.c | 8 ++-- 1 file changed, 2 insertions(+), 6 deletions(-) diff --git a/src/amd/common/ac_llvm_build.c b/src/amd/common/ac_llvm_build.c index 88ea289a121..253073e52fb 100644 --- a/src/amd/common/ac_llvm_build.c +++ b/src/amd

[Mesa-dev] [PATCH 01/11] ac: fix glc parameter use for new LLVM 8 typed buffer intrinsics

ac_build_llvm8_tbuffer_load() expects a boolean for glc. Fixes: 2cf5433b99f ("ac: use new LLVM 8 intrinsic when loading 16-bit values") Signed-off-by: Samuel Pitoiset --- src/amd/common/ac_llvm_build.c | 4 ++-- src/amd/common/ac_llvm_build.h | 2 +- src/amd/common/ac_nir_to_

[Mesa-dev] [PATCH 04/11] ac: add ac_build_buffer_store_format() helper

Similar to ac_build_buffer_load_format(). Signed-off-by: Samuel Pitoiset --- src/amd/common/ac_llvm_build.c | 100 src/amd/common/ac_llvm_build.h | 11 src/amd/common/ac_nir_to_llvm.c | 29 +++-- 3 files changed, 119 insertions(+), 21 deletions

[Mesa-dev] [PATCH 05/11] ac/nir: remove one useless check in visit_store_ssbo()

Trivial. Signed-off-by: Samuel Pitoiset --- src/amd/common/ac_nir_to_llvm.c | 9 +++-- 1 file changed, 3 insertions(+), 6 deletions(-) diff --git a/src/amd/common/ac_nir_to_llvm.c b/src/amd/common/ac_nir_to_llvm.c index c10a0cce16f..55d3ce90ce4 100644 --- a/src/amd/common/ac_nir_to_llvm.c

[Mesa-dev] [PATCH 08/11] ac/nir: use ac_build_buffer_store_dword() for SSBO store operations

Signed-off-by: Samuel Pitoiset --- src/amd/common/ac_nir_to_llvm.c | 23 +-- 1 file changed, 9 insertions(+), 14 deletions(-) diff --git a/src/amd/common/ac_nir_to_llvm.c b/src/amd/common/ac_nir_to_llvm.c index be0dca2d857..47a865de36f 100644 --- a/src/amd/common

[Mesa-dev] [PATCH 07/11] ac/nir: use ac_build_buffer_load() for SSBO load operations

Signed-off-by: Samuel Pitoiset --- src/amd/common/ac_nir_to_llvm.c | 35 ++--- 1 file changed, 6 insertions(+), 29 deletions(-) diff --git a/src/amd/common/ac_nir_to_llvm.c b/src/amd/common/ac_nir_to_llvm.c index 922f1063c79..be0dca2d857 100644 --- a/src/amd/common

[Mesa-dev] [PATCH 09/11] ac: use new LLVM 8 intrinsics in ac_build_buffer_load()

Signed-off-by: Samuel Pitoiset --- src/amd/common/ac_llvm_build.c | 8 1 file changed, 8 insertions(+) diff --git a/src/amd/common/ac_llvm_build.c b/src/amd/common/ac_llvm_build.c index bb8a470ae1d..7aec8154a76 100644 --- a/src/amd/common/ac_llvm_build.c +++ b/src/amd/common

[Mesa-dev] [PATCH 11/11] ac: use new LLVM 8 intrinsics in ac_build_buffer_store_dword()

New buffer intrinsics have a separate soffset parameter. Signed-off-by: Samuel Pitoiset --- src/amd/common/ac_llvm_build.c | 66 ++ 1 file changed, 26 insertions(+), 40 deletions(-) diff --git a/src/amd/common/ac_llvm_build.c b/src/amd/common/ac_llvm_build.c

[Mesa-dev] [PATCH 10/11] ac: use new LLVM 8 intrinsic when storing 16-bit values

Signed-off-by: Samuel Pitoiset --- src/amd/common/ac_llvm_build.c | 108 src/amd/common/ac_llvm_build.h | 26 src/amd/common/ac_nir_to_llvm.c | 26 ++-- 3 files changed, 139 insertions(+), 21 deletions(-) diff --git a/src/amd/common

Re: [Mesa-dev] [PATCH] radv: Fix driverUUID

On 3/12/19 7:11 PM, Emil Velikov wrote: On Tue, 12 Mar 2019 at 15:12, Samuel Pitoiset wrote: Reviewed-by: Samuel Pitoiset A fixes tag like below would be great ;-) Fixes: 14cad8786a8 ("radv: generate the same driver UUID as radeonsi") Reviewed-by: Emil Velikov Aside:

[Mesa-dev] [PATCH] radv: set the maximum number of IBs per submit to 192

This fixes random SteamVR corruption, see https://github.com/ValveSoftware/SteamVR-for-Linux/issues/181 Fixes: 4d30f2c6f42 ("radv/winsys: remove the max IBs per submit limit for the fallback path") Signed-off-by: Samuel Pitoiset --- src/amd/vulkan/radv_device.c

Re: [Mesa-dev] [PATCH 04/11] ac: add ac_build_buffer_store_format() helper

On 3/12/19 5:19 PM, Samuel Pitoiset wrote: Similar to ac_build_buffer_load_format(). Signed-off-by: Samuel Pitoiset --- src/amd/common/ac_llvm_build.c | 100 src/amd/common/ac_llvm_build.h | 11 src/amd/common/ac_nir_to_llvm.c | 29 +++-- 3

Re: [Mesa-dev] [PATCH 11/11] ac: use new LLVM 8 intrinsics in ac_build_buffer_store_dword()

:19 am, Samuel Pitoiset wrote: New buffer intrinsics have a separate soffset parameter. Signed-off-by: Samuel Pitoiset ---   src/amd/common/ac_llvm_build.c | 66 ++   1 file changed, 26 insertions(+), 40 deletions(-) diff --git a/src/amd/common/ac_llvm_build.c b/src

Re: [Mesa-dev] [PATCH 11/11] ac: use new LLVM 8 intrinsics in ac_build_buffer_store_dword()

On 3/13/19 9:01 AM, Samuel Pitoiset wrote: On 3/13/19 1:26 AM, Timothy Arceri wrote: This one causes 2000+ piglit tests to fail on radeonsi. For example: ./bin/shader_runner generated_tests/spec/arb_gpu_shader_fp64/execution/conversion/geom-conversion-explicit-bool-double.shader_test -auto

[Mesa-dev] [PATCH] ac: do not force enable IDXEN for 16-bit SSBO loads

The struct version enables IDXEN, while the raw one disables it. When vindex is unused, the raw version is enough. Signed-off-by: Samuel Pitoiset --- src/amd/common/ac_llvm_build.c | 4 +++- 1 file changed, 3 insertions(+), 1 deletion(-) diff --git a/src/amd/common/ac_llvm_build.c b/src/amd

[Mesa-dev] [PATCH v2 01/12] ac: do not force enable IDXEN for 16-bit SSBO loads

The struct version enables IDXEN, while the raw one disables it. When vindex is unused, the raw version is enough. Signed-off-by: Samuel Pitoiset --- src/amd/common/ac_llvm_build.c | 4 +++- 1 file changed, 3 insertions(+), 1 deletion(-) diff --git a/src/amd/common/ac_llvm_build.c b/src/amd

[Mesa-dev] [PATCH v2 00/12] ac: use LLVM 8 buffer intrinsics everywhere

Hi, This small series makes use of new LLVM 8 buffer intrinsics. No CTS regressions on GFX8 with LLVM 7, 8 and master. V2: fix use of IDXEN for GFX9 Please review, Thanks! Samuel Pitoiset (12): ac: do not force enable IDXEN for 16-bit SSBO loads ac: fix glc parameter use for new LLVM 8

[Mesa-dev] [PATCH v2 10/12] ac: use new LLVM 8 intrinsics in ac_build_buffer_load()

v2: - do not force enable IDXEN when unecessary Signed-off-by: Samuel Pitoiset --- src/amd/common/ac_llvm_build.c | 10 ++ 1 file changed, 10 insertions(+) diff --git a/src/amd/common/ac_llvm_build.c b/src/amd/common/ac_llvm_build.c index 5d52d93d25a..8d5682f6f7a 100644 --- a/src/amd

[Mesa-dev] [PATCH v2 06/12] ac/nir: remove one useless check in visit_store_ssbo()

Trivial. Signed-off-by: Samuel Pitoiset --- src/amd/common/ac_nir_to_llvm.c | 9 +++-- 1 file changed, 3 insertions(+), 6 deletions(-) diff --git a/src/amd/common/ac_nir_to_llvm.c b/src/amd/common/ac_nir_to_llvm.c index c10a0cce16f..55d3ce90ce4 100644 --- a/src/amd/common/ac_nir_to_llvm.c

[Mesa-dev] [PATCH v2 02/12] ac: fix glc parameter use for new LLVM 8 typed buffer intrinsics

ac_build_llvm8_tbuffer_load() expects a boolean for glc. Fixes: 2cf5433b99f ("ac: use new LLVM 8 intrinsic when loading 16-bit values") Signed-off-by: Samuel Pitoiset --- src/amd/common/ac_llvm_build.c | 4 ++-- src/amd/common/ac_llvm_build.h | 2 +- src/amd/common/ac_nir_to_

[Mesa-dev] [PATCH v2 05/12] ac: add ac_build_buffer_store_format() helper

Similar to ac_build_buffer_load_format(). v2: - fix out of bounds access Signed-off-by: Samuel Pitoiset --- src/amd/common/ac_llvm_build.c | 100 src/amd/common/ac_llvm_build.h | 11 src/amd/common/ac_nir_to_llvm.c | 29 +++-- 3 files changed, 119

<    3   4   5   6   7   8   9   10   11   12   >